entorin

Overview

What you get from Entorin, mapped to the harness pains every agent stack reinvents.

Entorin is a control plane for agent harnesses, not a platform. Whatever you already use to drive agents — Claude Agent SDK, Codex CLI / Codex SDK, a hand-written while loop, LangGraph — keeps its shape. Entorin slides underneath and supplies the harness layer: observability, budget, sandbox, auth, HITL.

A standardised lower layer that every harness reinvents, factored out so you don’t.

Layer model

Agent runtime

LangChain/LangGraph

Claude Agent SDK

OpenAI Agents SDK

Agent framework

LangChain/LangGraph

CrewAI

Google ADK

LlamaIndex

Agent harness

Claude Code

Codex CLI

Hermes Agent

Entorin — control plane

Entorin plugs into every layer. The harness owns the workflow; Entorin owns the cross-cutting infrastructure every harness otherwise rebuilds.

What you get, by pain point

PainWhat the integration gives you
ObservabilityOne OTel trace per run. Every LLM call, tool call, agent invocation, sandbox exec, and checkpoint round-trip is a span carrying entorin.run_id, entorin.principal_id, tokens, cost.
Framework lock-inThe bare-loop reference shows that entorin itself never asks you to subclass anything or build a DAG. A 50-line Python while loop inherits the full harness.
Testing & evalsSaved traces become regression fixtures. entorin.replay ships a TraceRecorder and a small set of invariant checks (assert_calls_paired, assert_run_lifecycle, assert_budget_within_cap).

Install

uv add entorin
# Optional extras:
uv add 'entorin[mcp]'    # MCP transport for the tool wrapper
uv add 'entorin[http]'   # FastAPI-backed HITL checkpoint transport